Individuals with social anxiety disorder (SAD) experience heightened anxiety in in-person communication, particularly when they perceive negative evaluations from their conversation partner based on eye gaze and facial expressions. To address this issue, this study developed an augmented reality (AR) system for in-person conversations, capable of concealing the presence and masking the facial emotions of the conversation partner. An online survey was conducted using Amazon Mechanical Turk (n = 130) to examine the visual efects that individuals with SAD find anxiety-inducing during in-person conversations. 91 out of them have a tendency for SAD. For the aspect of "ease to talk", individuals with SAD achieved significantly higher scores for the visual efects involving a smiley face or anime avatar overlaid onto the conversation partner, as opposed to the remaining visual efects (p < .05). Based on these findings, we developed two prototypes wherein an AR head-mounted display (HMD) overlays either a smiley face or an anime avatar onto the conversation partner during an in-person conversation. A small-scale user study was conducted to investigate participant's ability to interact with a conversation partner, using these visual efects, as well as their subjective experiences. The results were further analyzed through user interviews. The social anxiety levels of participants were measured using the Liebowitz Social Anxiety Scale. Participants with SAD exhibited a preference for communicating with an anime avatar rather than real individuals, while participants without SAD preferred real-person interactions. These findings inform future research, wherein the objective (heart rate, eye tracking, self-disclosure) and quantitative data (questionnaires) will be utilized to evaluate anxiety levels during in-person conversations.
resistant to change. Furthermore, barriers to seeking treatment involve an inadequate acknowledgement of the condition by healthcare professionals, apprehension regarding negative evaluation by healthcare providers, and a lack of awareness regarding the availability of effective therapeutic treatment across various regions [3].
A theoretical framework [4] describes the cognitive processes involved in the genesis of social anxiety, and clarifies the psychological mechanisms. Individuals with SAD tend to engage in a cognitive distortion wherein they incorrectly perceive others as evaluating them negatively, despite the absence of explicit negative evaluations. Additionally, in comparison to individuals without SAD, individuals with SAD exhibit a heightened tendency to selectively attend to threatening stimuli, particularly in social contexts. Within such social situations, cues indicative of potential negative evaluations from others, such as eye gaze and facial expressions, serve as salient threats. Consequently, individuals with SAD experience heightened anxiety when subjected to gaze from others, particularly when accompanied by neutral or negative facial expressions (e.g., anger, fear) [5]. Notably, individuals with SAD display a greater degree of comfort when
bia, is one of the most prevalent mental disorders in the world [1]. SAD induces profound anxiety in response to social circumstances such as face-to-face communication and public speaking. Consequently, individuals with SAD often encounter a heightened likelihood of lower education, lower occupational status, and diminished income when compared to individuals without SAD [2].
Psychological and pharmacological interventions are employed to address SAD. However, the rate of seeking treatment is low because individuals with SAD perceive that social anxiety is intrinsic to their personality and of eye gaze and facial emotion as triggers for their social ofers individuals with SAD the opportunity to receive anxiety within social settings [6]. treatment remotely without the need for in-person coun
Although communication without visual cues such as selling. Andersson [10] devised ICBT by adapting estabtext message promotes a relative reduction in anxiety lished CBT protocols and incorporating group exposure experienced by individuals with SAD, numerous social sessions. Participants in their study engaged in therapy circumstances need in-person interactions, such as seek- delivered via the Internet, consisting of learning materiing to counsel in a hospital or dealing with administrative als and exercises. These exercises prompted participants tasks at a city hall. However, few research aims to aid to clarify their thoughts and respond to specific inquiries, individuals with SAD for an in-person conversation. facilitating therapist evaluation of material assimilation.
Augmented reality (AR) is a transformative technol- Furthermore, participants received personalized feedback ogy that enables individuals to perceive the real world from therapists via email. CBT and ICBT are grounded in supplemented with virtual objects. AR has brought about the theoretical understanding that anxiety in SAD derives enhanced convenience and futuristic experiences, rang- from cognitive distortions. Their therapeutic objective ing from visualizing virtual constructions within real- is the reconstruction of their cognition. However, an world environments to supporting surgeons through in- alternative perspective suggests that anxiety in SAD can formation navigation for their surgery [7]. However, the also arise from a lack of social interaction skills. Accordadoption of AR technology to support individuals with ing to this viewpoint, individuals with SAD can mitigate disabilities or impairments remains limited, for example, their symptoms by practising social skills. By addressthe development of an AR system that enables individ- ing both cognitive distortions and a lack of social skills, uals with colour blindness to perceive real-time scenes comprehensive treatment approaches can ofer a more through a filtered image [8]. efective approach to reducing social anxiety symptoms
To address this research gap, our study aims to miti- and enhancing overall social functioning [11]. gate the anxiety experienced by individuals with SAD Social skills training (SST) is a therapeutic approach during in-person conversations. We aim to accomplish that aims at enhancing social functioning by practising this by designing an AR system capable of modifying or social skills. It has shown efectiveness in addressing varconcealing the eye and facial features of others, which ious mental disabilities, including SAD, schizophrenia, have been identified as triggers for social anxiety. We and autism spectrum disorder. SST is typically conducted aspire to provide an approach for supporting individu- by psychological specialists and involves structured exerals with SAD to reduce their anxiety about in-person cises that incorporate real-life social interactions, such as conversation in their daily life. making eye contact, asking open-ended questions, and taking turns. Role-playing exercises are utilized in SST [12]. Virtual reality (VR) has been employed in the con2. Related works text of social skills training, specifically for individuals with schizophrenia [13]. These interventions utilize VR The treatment and training for SAD have traditionally to create immersive environments for practising converbeen administered by human professionals, specifically sation skills, assertiveness, and emotional expression. By medical specialists. These days, a number of research has engaging in simulated social interactions within a conexplored the utilization of computer technologies such trolled virtual setting, individuals can develop their social as the Internet and virtual reality (VR). skills in a safe and supportive environment. However,
Cognitive behaviour therapy (CBT) is an established it should be noted that individuals with SAD may pospsychological treatment employed in medical institutions sess social skills but struggle to demonstrate them due to for SAD and has demonstrated efectiveness [ 9]. Through anxiety or negative thoughts regarding their behaviour. structured exercises facilitated by a therapist, patients It remains unclear whether SST is essential for every acquire coping skills and learn to modify their cogni- individual with SAD [11]. tion, problematic emotions, and behaviours. Despite In order to avoid the challenges of in-person commuthe overall eficacy of CBT, the utilization of this treat- nication, text messages, including email and chat appliment remains limited. This is because individuals with cations, are feasible alternatives. Pierce [6] found that SAD perceive their condition as their personality and individuals with SAD tend to prefer text messaging over apprehensions about negative evaluations from health- face-to-face communication. Kang et al. [14] investicare professionals prevent them from seeking counseling. gated the disclosure patterns of individuals with SAD Additionally, there exists a lack of information regard- in interactions with 3D avatars compared to real indiing the availability of efective treatment in various re- viduals through video calls. They found that individuals gions [3]. Internet-delivered self-help therapy, known with SAD were more likely to disclose themselves to as Internet-based cognitive behaviour therapy (ICBT) 3D avatars than to conversation partners during video holds the potential for addressing the challenges. ICBT calls. While Kang et al. focused on altering only the appearance of the conversation partner, Ichino et al. [15] conducted a study exploring self-disclosure by modifying both the appearance of the subject and the conversation partner. They compared three conditions: avatars with high similarity to the participant in virtual reality (VR), avatars without similarity to the participant in VR, and unaltered video calls. The results demonstrated that participants exhibited the highest level of self-disclosure with avatars that lacked similarity to them in VR compared to the other conditions. Although this research did not specifically target individuals with SAD, it suggests that employing visual efects and mediums such as video or VR in face-to-face communication may potentially encourage individuals to engage more in conversation.
Building upon the limitations of existing communication tools, a system which can support individuals with SAD for in-person interaction is essential. the level of anxiety they experienced in relation to the conversation partner and the ease of communication in each condition.
The proposed AR system is designed specifically for 3. Main idea dyadic in-person conversation scenarios, where individuals with SAD can utilize the system during an in-person 3.1. Causes of anxiety for in-person conversation. This approach is particularly relevant in situations where the conversation partner does not excommunication plicitly evaluate individuals with SAD negatively (e.g. Individuals with SAD often experience heightened anxi- asking for directions, self-introductions, and engaging ety in in-person conversation situations due to negative in small talk with strangers). The current adaptation of interpretations. Research [5] has demonstrated that fac- AR HMDs into our daily lives is limited. However, retors such as eye gaze and facial emotion can trigger anxi- cent advancements in technology have resulted in the ety in individuals with SAD. Once they perceive a nega- development of smaller AR HMDs, such as the XReal Air tive evaluation from the conversation partner, their atten- [17], which is similar to sunglasses. This indicates the tion becomes focused on their own physiological arousal, potential that AR HMDs will blend into our daily life. which may manifest as increased heart rate, sweating, or blushing. This heightened self-awareness intensifies 3.3. Research aims their anxiety and leads to avoidance behaviours. Consequently, they feel anxiety about engaging in conversation.
individuals with SAD when engaging in in-person conversation by developing an AR system. As an initial phase, we aim to investigate the types of visual efects that can reduce anxiety levels in individuals with SAD. Furthermore, we seek to explore their ability to efectively communicate in situations where visual efects are overlaid onto their conversation partner. This is because overlaying of visual efects on conversation partners is not a common occurrence in real-life interactions and may elicit unfamiliar sensations or perceptions.
This study proposes an AR approach to address the causes of anxiety in in-person communication by manipulating the presence and facial emotions of the conversation partner. The method involves individuals with SAD wearing an AR head-mounted display (AR HMD) during in-person interactions. AR HMD rendered virtual objects onto the conversation partner (Fig. 1). To investigate the efective- 4. Pre-survey ness of diferent visual efects in changing or hiding the conversation partner, an online survey was conducted us- In order to find out which specific visual efects are most ing Amazon Mechanical Turk [16]. Considering that eye helpful in supporting ease of conversation and reducing gaze and facial emotions are key triggers of anxiety, eight anxiety among individuals with SAD, we conducted an visual efects were prepared. The participants of the sur- online survey via Amazon Mechanical Turk [18]. vey were presented with nine conditions, including eight variations of visual efects and one condition without any visual efect. The participants were asked to evaluate
For our study, we prepared eight diferent visual efects that could be overlaid onto the appearance of the conversation partner (Fig. 2). These visual efects were specifically designed to address the sources of anxiety such as eye gaze and facial expressions experienced by individuals with SAD. The visual efects included: (a) Nothing overlaid (control condition): No visual efect applied, (b) Mosaic pattern eyes overlaid: The conversation partner’s eyes are covered with a mosaic pattern, (c) White square overlaid: A white square is overlaid onto the conversation partner’s eyes, (d) Anime eyes overlaid: The conversation partner’s eyes are overlaid with anime-style eyes, (e) Smiley face overlaid: A smiley face is overlaid with the conversation partner’s face, (f) Anime avatar face overlaid: The conversation partner’s face is overlaid with an anime-style avatar face, (g) Realistic avatar face overlaid: The conversation partner’s face is overlaid with a realistic avatar face, (h) Anime avatar overlaid: The entire body of the conversation partner is overlaid with an anime-style avatar, (i) Realistic avatar overlaid: The entire body of the conversation partner is overlaid with a realistic avatar. We chose a white square as one of the visual efects because it appears nearly opaque when viewed through devices such as Microsoft Hololens2 [19] therefore it can efectively cover the conversation partner’s eyes. The use of a mosaic pattern is a common method of concealing the eyes. Additionally, research by Nowak and Raugh [20] has shown that anthropomorphic avatars are perceived as attractive and credible. Therefore, we incorporated anime-style eyes, faces, and full bodies into our visual efects options. Realistic avatars were also included as they are more lifelike than anime-style avatars but not as realistic as actual people. Kang et al. [14] demonstrated that individuals with SAD were more willing to disclose personal information to a realistic avatar compared to actual people during video calls. However, realistic avatars may trigger feelings of discomfort due to the possibility of encountering the uncanny valley phenomenon [21], where people experience unease when an avatar becomes almost indistinguishable from a real human but still falls short of complete realism.
We collected responses from a total of 130 participants through Amazon Mechanical Turk. Each participant received a compensation of $1 for completing all the survey questions. To ensure that our analysis focused specifically on individuals with SAD, we filtered the data using the Liebowitz Social Anxiety Scale (LSAS) [22]. Out of the 130 participants, 91 individuals (42 females, 49 males) scored above the threshold for SAD on the LSAS. The age range of these socially anxious participants was between 18 and 69 years. In terms of geographical distribution, 53 participants resided in North America, 34 in Asia, two in South America, and one each in Europe and Africa. Regarding the participants’ self-identified ethnic groups, 46 described themselves as White, 34 as Asian or Pacific Islander, four as Latino or Hispanic, four as Black or African American, one as Native American or American Indian, and one participant chose "Other". One participant did not provide an answer to this question.
about demographic questions such as gender, age, selfidentified ethnic groups, and place of residence. The second section is LSAS to filter the people with SAD. The third section is Gaze Anxiety Scale [23] to measure the level of gaze avoidance. The fourth and fifth sections are image evaluation and comparisons among nine conditions. We showed nine visual efects and the introduction "You are wearing AR glasses and try to talk with strangers for asking for directions. In the AR environment, you can overlay a visual efect on the stranger. Please answer the following questions per condition (Fig. 2)". The fourth section is image evaluation. Participants evaluated the nine visual efects individually. Using a seven-point Likert scale (ranging from 1: strongly disagree to 7: strongly agree), participants rated statements such as "I feel that this person would be easy to talk to" and "I feel the anxiety to talk to this person". The fifth section is image comparison. Participants chose the best visual efect from all conditions for specific questions like "Which person do you feel is the easiest to talk to?" and "Which person do you feel the most anxiety to talk to?" Participants selected their preferred visual efect based on their perception of ease or anxiety in communication.
The statistical analysis was conducted using nonparametric tests due to the non-normal distribution of the data, as confirmed by the Shapiro-Wilk test. Friedman’s test was used to determine if there were significant diferences among the multiple related groups, and the results indicated a significant diference among the conditions ( < .05). To compare paired observations, the Wilcoxon signed-rank test was employed, with a significance level set at = .05. The statistical analyses were performed using IBM SPSS Statistics software [25].
Regarding the image comparison section, the results for the question "Which person do you feel is the easiest to talk to?" showed that (e) was selected by more than 30 out of the 91 participants (Fig. 3). Additionally, for the question "Which person do you feel the most anxiety to 30 ) n ( p i c i t s tan20 r aP10 0 d() a() g() overlaid (control condition), (b) mosaic pattern eyes overlaid (c) white square overlaid (d) anime eyes overlaid, (e) smiley face overlaid, (f) anime avatar’s face overlaid, (g) realistic avatar’s face overlaid, (h) anime avatar’s body overlaid, (i) realistic avatar’s body overlaid. The image of the conversation partner is a royalty-free image from photo AC [24].
(a) (b) (c) (d) (e) (f) (g) (h) (i)
Conditions condition for "Which person do you feel easiest to talk to". talk to?", Figure 4 indicates that (a) was selected by more than 20 out of the 91 participants. than the rest of the visual efects for "I feel that this person would be easy to talk to". There is also a significant diference between (e) and the rest of the visual efects(Fig.
scored significantly higher than most of the rest of the visual efects. (e), (f) and (h) for "I feel the anxiety to talk to this person"
Conditions condition for "Which person do you feel the anxiety to talk to is the most?" (a) (b) (c) (d) (f) (g) (h)
(i) (e)
Contidions to range from 1 to 7 for the question "I feel that this person would be easy to talk to" in the image comparison section. Mean values and standard deviation of "I feel easy to talk to this person" and "I feel the anxiety to talk to this person" based on likert scale to a range from 1 to 7.
Feel easy to talk
Feel anxiety to talk () 4.48 (1.486) 3.95 (1.493) 3.86 (1.488) 3.79 (1.567) 5.07 (1.467) 4.37 (1.644) 3.59 (1.619) 4.57 (1.674) 3.80 (1.701) () 4.15 (1.653) 4.05 (1.566) 4.02 (1.535) 4.00 (1.606) 3.58 (1.713) 3.58 (1.720) 4.27 (1.745) 3.47 (1.715) 3.98 (1.713) avatar. Participants may have found it strange to interact with the conversation partner whose appearance was overlaid with visual efects, as this is not commonly experienced in real-world interactions. The visual efects were designed to cover the eyes, considering that individuals with SAD tend to avoid direct gaze. However, participants scored significantly lower for visual efects (b), (c), and (d) in terms of feeling that the person would be easy to talk to compared to (e). This suggests that simply hiding the eyes of a conversation partner may not necessarily make individuals with SAD less anxious in an in-person conversation. Furthermore, visual efects (g) and (i), which feature a realistic avatar only on the face, also scored significantly lower than (e). This could be attributed to the potential uncanny valley efect, where discomfort arises when an avatar appears too realistic [21]. (a) (b) (c) (d) (f) (g) (h)
(i) (e)
Contidions
The findings of the study indicate that participants tend to feel more at ease and less anxious when interacting 5. Prototype with a person who is smiling. This supports previous research that suggests individuals with SAD experience 5.1. Design anxiety related to the facial emotions of others [5]. Additionally, participants reported feeling more comfortable Our AR system is designed for supporting dyadic intalking to a person whose full body is covered by an person conversations for individuals with SAD. This sysanime avatar. This aligns with the notion that anthropo- tem involves individuals with SAD wearing an AR HMD morphic avatars are perceived as attractive and credible which visual efects are rendered onto the conversation [20]. Although the proportion of participants who chose partner(Fig. 1). Based on insights gained from a prethe control condition is lower compared to the smiley liminary survey, we have developed a prototype that image and full body covered by the anime avatar, it is incorporates two distinct visual efects. Fig. 7 is the still comparable to the full body covered by the anime view through the AR HMD during the dyadic in-person
conversation without any visual efect. The first efect 6.1. Participants involves displaying a 2D image of the conversation partner’s smiley face, aimed at altering their facial emotion The study involved five male students from the infor(smiley face condition)(Fig. 8). However, the static smiley mation science department (Nara Institute of Science face image may seem incongruous when the conversa- and Technology), who voluntarily participated. All partion partner is talking. To address this, 2D image of the ticipants were Japanese and proficient in the Japanese conversation partner changes to an image where their language. Their ages ranged from 22 to 24 years, with mouth appears to open when they start talking and re- a mean age of = 22.6 and a standard deviation of verts back to the smiley face when they stop speaking = 0.89. To measure their levels of social anxiety, the (Fig. 8). In the second visual efect, we employ a 3D anime Japanese version of the Liebowitz Social Anxiety Scale avatar that matches the gender of the conversation part- (LSAS-J) [22] was administered, resulting in a mean score ner, efectively concealing their physical presence (anime of = 48.6 and a standard deviation of = 21.93. avatar condition)(Fig. 9). Its mouth movements are syn- The threshold for SAD on the LSAS-J is above 30 (Table. chronized with the ones of the conversation partner (Fig. 4). 9). Additionally, based on the findings of Takashima et al. [26], which highlighted the influence of avatar blinking on subject impressions, we have also synchronized Table 4 the blink rate of the avatar to create a more natural and Age and LSAS score of participants of user study friendly impression.
6.3.2. Smiley face condition The user study began with a brief explanation of the Despite the participants’ overall inability to perceive the task, during which participants were given the opportu- eye gaze and facial emotions of the conversation partnity to ask any questions related to the study. Following ner, none of them expressed a preference for the smiley this, participants completed the LSAS-J questionnaire. face condition compared to the other conditions. One Once the questionnaire was completed, the experimental participant specifically mentioned feeling strange about sessions commenced. Participants wore the HoloLens2 the fact that only the face of the conversation partner device and calibrated their eye movements. They then was represented by a 2D image. This suggests that the engaged in conversations with a conversation partner use of a smiley face overlay did not efectively meet the who was a well-trained student from the information participants’ expectations or provide a sense of comfort. science department (Nara Institute of Science and Tech- Furthermore, some participants reported that the smiley nology). Although participants had some familiarity with face condition did not completely conceal the conversathe conversation partner, they were not extensively famil- tion partner’s actual facial features. Participants without iar with each other. The conversation partner maintained SAD or with lower social anxiety expressed a desire to a neutral facial expression and engaged in passive, non- see the actual facial emotions of the conversation partner. negative communication with the participants. The study They believed that observing the genuine facial expresconsisted of three conditions: anime avatar overlaid, smi- sions would provide them with valuable cues about how ley face overlaid, and no overlay (control condition). The the conversation partner was reacting to their interacorder of these conditions was randomly assigned to each tions. One participant with generalized social anxiety participant. Three topics were provided for discussion, reported feeling a sense of safety in the smiley face conwith each topic randomly assigned to one of the condi- dition compared to the condition where nothing was tions. The topics included thoughts about trips abroad, overlaid onto the conversation partner’s face. high school life, and experiences at a hair salon. Upon completion of the user study, participants were asked 6.3.3. Control condition to answer a final questionnaire to indicate which condition they felt easy or anxious. Additionally, they participated in interviews to provide further insights on the user study.
Participants with SAD exhibited a preference against the control condition. They expressed that being able to see the face of the conversation partner induced anxiety, as they were concerned about the partner’s thoughts and judgments. Moreover, some participants reported heightened anxiety due to their attention being drawn to the eye gaze of the conversation partner. In contrast, participants without SAD or with lower levels of SAD preferred the control condition. They mentioned that the anime avatar and smiley face conditions, which obscured the facial emotion of the conversation partner, caused them to worry about the partner’s reactions and responses.
6.3.1. Anime avatar condition The results of the user study indicated that all participants 6.4. Discussion reported an inability to see the eye gaze and facial expressions of the conversation partner. Participants with We hypothesized that participants with SAD would exSAD preferred the conditions where the conversation hibit a preference for conditions where visual overlays partner was overlaid with an anime avatar. They found are present, as these overlays can reduce sources of anxit more comfortable and less anxiety-inducing compared iety [5]. Consistent with our expectations, participants to the control condition. Participants without SAD or with SAD preferred the anime avatar condition, which efwith lower social anxiety, however, preferred the con- fectively concealed the facial emotion of the conversation trol condition. They reported feeling more at ease and partner, compared to the control condition. This suggests natural when communicating without the presence of that individuals with SAD are capable of engaging in conan avatar overlay. Participants who preferred the anime versation with a conversation partner whose presence is avatar condition expressed that they did not feel strange obscured by an anime avatar. Interestingly, despite their or uncomfortable communicating with the avatar. They social anxiety, participants did not prefer the smiley face reported that the experience is similar to interacting with condition. This can be attributed to the perceived unnatvirtual characters such as Vtubers or anime characters uralness associated with the overlaid 2D image, which with whom they were familiar. surpasses the anxiety deriving from the eye gaze and facial emotion of the conversation partner. As for only one participant with generalized SAD, the absence of per- online survey to investigate the specific visual efects ceived eye gaze and facial emotions, combined with the that individuals with SAD perceive as reducing their anxpresence of a smiling 2D image, contributed to a feeling iety. Our findings revealed that facial expressions and of security. the visual appearance of these efects serve as cues for
Conversely, participants without SAD or with lower reducing anxiety when talking with people. Additionlevels of SAD expressed a preference for the control con- ally, we conducted a preliminary user study to explore dition. They experienced anxiety due to the inability to how individuals with SAD engage and experience comperceive the genuine emotions and reactions of the con- munication when an anime avatar or a 2D image of a versation partner. These findings highlight the divergent smiley face is overlaid onto their conversation partner. sources of anxiety experienced by individuals with SAD Qualitative research methods, including user interviews, and those without SAD. were employed for evaluation. Participants indicated that
Participants were also asked about the individuals communication with a conversation partner concealed with whom they experienced anxiety while engaging by an anime avatar is possible. Facial expressions of the in conversation and the underlying reasons for their anx- conversation partner play a significant role in influenciety. All participants reported feeling anxiety when talk- ing the anxiety levels of individuals with SAD, however, ing with individuals in positions of high authority, such the smiley face condition did not yield positive outcomes. as professors, bosses, or seniors. The reason cited for Therefore, further improvements are required in our futhis anxiety was the concern of being negatively evalu- ture work. Additionally, the study identified contrasting ated by them. Participants expressed worries that their experiences between individuals with SAD and those knowledge would be deemed insuficient, that they might without SAD, which presents an intriguing finding dediscuss topics inaccurately, or that their conversations serving of further exploration. would be perceived as a waste of time by individuals in This study is subject to certain limitations. Firstly, the positions of high authority. Furthermore, some partici- sample size is small and we did not incorporate quantipants mentioned that their anxiety levels were mitigated tative data, such as biological indicators such as heart when individuals in positions of high authority exhibited rate and eye movement, and subjective questionnaires friendly and casual behaviour. These findings align with such as assessments of anxiety pertaining to verbal comthe social phobia model proposed by Rapee and Heim- munication, in our evaluation. Secondly, participants berg [4], which assumes that individuals with SAD hold were not encountering their conversation partners for a strong belief that they are being negatively evaluated the first time, which may have resulted in lower levels of by others, even in the absence of explicit evaluation. The anxiety during the conversation. Furthermore, the user anime avatar is lacking any cues suggesting high author- study was exclusively conducted with male students, and ity. Consequently, participants may not perceive inter- the sample size was inadequate. We intend to incorpoacting with individuals in the position of high authority, rate quantitative measurements to assess anxiety levels therefore our system may support them. Conversely, in during interactions with strangers. the smiley face condition, participants may still recognize In conclusion, the outcomes of this study bear signifthe conversation partner as a high-authority individual. icant implications for the design of technological interFuture investigations should consider the potential influ- ventions for SAD and the enhancement of individuals’ ence of vocal cues on participants’ anxiety levels when quality of life. For instance, individuals with SAD can talking with high-authority individuals. actively engage in education and broaden their career
Several individuals experiencing social anxiety have prospects. reported feelings of apprehension when engaging in conversations with unfamiliar individuals and those of the opposite gender. We employed an anime avatar charac- 8. Acknowledgments terized by a gender matching that of the conversation partner. While disparities in vocal expression persist, our AR system holds promise in reducing conversational anxiety when the anime avatar’s gender aligns with that of the user employing our AR system.
JPMJCR19A, Japan.
We would like to acknowledge the guidance provided by Senior Lecturer Mitsuhiro Uratani from Nara Medical Universiy, Assistant Professor Hiroki Tanaka and Professor Satoshi Nakamura from Nara Institute of Science and Technology.
153–178. URL: https://onlinelibrary.wiley.com/doi/ abs/10.1111/j.1083-6101.2006.tb00308.x. doi:https: //doi.org/10.1111/j.1083-6101.2006. tb00308.x. [21] M. Mori, K. F. MacDorman, N. Kageki, The uncanny valley [from the field], IEEE Robotics Automation Magazine 19 (2012) 98–100. doi:10.1109/ MRA.2012.2192811. [22] R. Heimberg, K. Horner, H. Juster, S. Safren, E. Brown, F. Schneier, M. Liebowitz, Psychometric properties of the liebowitz social anxiety scale, Psychological medicine 29 (1999) 199–212. doi:10.1017/S0033291798007879. [23] F. R. Schneier, T. L. Rodebaugh, C. Blanco, H. Lewin, M. R. Liebowitz, Fear and avoidance of eye contact in social anxiety disorder, Comprehensive Psychiatry 52 (2011) 81– 87. URL: https://www.sciencedirect.com/science/ article/pii/S0010440X10000398. doi:https://doi. org/10.1016/j.comppsych.2010.04.006. [24] ACworks Co.,Ltd. , photo ac, 2023. https://en.
photo-ac.com/. [25] IBM, IBM SPSS Statistics, 2023. https://www.ibm.
com/products/spss-statistics. [26] K. Takashima, Y. Omori, Y. Yoshimoto, Y. Itoh, Y. Kitamura, F. Kishino, Efects of avatar’s blinking animation on person impressions, GI ’08, Canadian Information Processing Society, CAN, 2008, p. 169–176. [27] Unity Software, Unity, 2023. https://unity.com/. [28] EnoxSoftware, HoloLensWithDlibFaceLandmarkDetectorExample, 2023. https://github.com/EnoxSoftware/ HoloLensWithDlibFaceLandmarkDetectorExample/ blob/master/Assets/ HoloLensWithDlibFaceLandmarkDetectorExample/ HLARHeadExample/HLARHeadExample.cs. [29] pixiv Inc., Vroid, 2023. https://vroid.com/en.